I build retrieval and LLM systems that actually ship. Currently at Asurion replacing a 3-year-old semantic search with a modular RAG pipeline serving 10,000+ daily queries, while finishing my M.S. in Data Science & AI at USF.
I'm an AI Engineer based in San Francisco. My day-to-day is RAG plumbing: ingestion, chunking, embeddings, retrieval, evals — making sure the thing that comes out the other end is actually better than what was there before.
Right now I'm at Asurion, where I built a modular RAG system that replaces a three-year-old semantic search serving 10,000+ daily customer queries. Before that I shipped a React Native app at StudyStudio.ai, an NLP matching system at USF's Data Institute that's still in production, and time-series pipelines for residential energy research at Frontier Energy.
Outside of work I ran USF's 100+ member rock climbing club for two years. I like things that are precise, durable, and slightly understated.
Multi-service GCP system that triages support tickets by predicted priority with a fine-tuned DistilBERT classifier. FastAPI backend on Cloud Run with Firebase auth and per-user rate limits, Cloud SQL Postgres scoped by uid, Vertex AI for training and the model registry.
From-scratch GPT-2 small in PyTorch (~124M params): BPE tokenizer with a sliding-window dataloader, multi-head causal self-attention, the full transformer stack, and a pretraining loop that samples generations mid-training. Each file maps to a stage of Sebastian Raschka's Build a Large Language Model.
A Claude Code skill and matching sub-agent that turn the main agent into a coordinator. Decomposes a plan into waves, spawns implementer sub-agents in isolated git worktrees, runs them in parallel where safe, and stays out of the way between checkpoints.
U-Net CNN for image segmentation that automates cell counting for biomedical research. Mean error of 1.4 cells, beating the 3-cell target by 2x and replacing manual counting.